rusti: duplicate definitions are accepted but break all further interaction

[ben0x539]

sample session:

$ rusti
WARNING: The Rust REPL is experimental and may be
unstable. If you encounter problems, please use the
compiler instead.
rusti> fn f() {}
()
rusti> fn f() {}
()
rusti> 42;
<anon>:35:0: 35:10 error: duplicate definition of value f
<anon>:35 fn f() { }
          ^~~~~~~~~~
<anon>:33:0: 33:10 note: first definition of value f here:
<anon>:33 fn f() { }
          ^~~~~~~~~~
error: aborting due to previous error
rust: task failed at 'explicit failure', /home/ben/code/repos/rust/src/libsyntax/diagnostic.rs:99

[jbclements]

Hang on... it's not at all clear to me that this should be fixed by allowing duplicate definitions. What's the semantic model for redefinitions? Scoping works differently for items, and it seems like we dodge a lot of questions by disallowing redefinitions here. In fact, it's not clear that the OP was actually asking for duplicate definitions, but simply for a better/earlier error message. Can I nominate this for WONTFIX ? Maybe I officially nominate it for not-a-bug / well-defined.

[ben0x539]

I hoped we'd at least reject the duplicate definition right away rather than accepting it and then rejecting all further input, it doesn't seem very productive to brick the whole session but still accept input.

I'd also say that redefinitions are really helpful for exploratory programming in the REPL. So even if this change isn't optimal yet (I haven't tried it), it's probably better than the current behavior?

[jbclements]

I certainly agree that it's far better to give an error message on the redefinition.

[alexcrichton]

Closed by #7070